How will accountants learn new skills when AI does the work?

One of the first things Carl Mayes, CPA, said he did when he started his career as an auditor 20 years ago was to vouch: He examined transactions in a company’s accounting records and compared them to supporting documentary evidence, like invoices and receipts, to ensure they were genuine, accurate, and properly recorded.

Today, much of that time-consuming and repetitive work is being automated.

Commercially available AI-powered platforms for audit and advisory firms “have vouching tools that will do that for you and that is going to save so much time,” said Mayes, vice president—CPA Candidate Quality & Competency at the AICPA. These tools can accelerate vouching and related audit procedures if staff know what information to provide the tool, how it ties to audit assertions, and how to critically evaluate AI-generated work.

Across the profession, automation and artificial intelligence (AI) are taking over much of the repetitive, low-risk work young accountants used to do to learn about systems, controls, and professional skepticism. That leaves the profession with a fundamental question: How do you train accountants when “the training work” is disappearing?

As a replacement, leaders from the AICPA, academia, and firms suggest a system that teaches conceptual mastery, human judgment, technological fluency, and adaptability.

As Mayes put it, the profession must pivot from “doing” everything to “supervising” more.

“When it’s a bot doing it, you need somebody to supervise that bot,” he said. “We can’t just accept that the bot has done it right. For us to supervise the bot, we have to understand the underlying procedure, even if we’re never going to do the procedure ourselves.”

This shift from executor to evaluator requires entirely different training.

FROM MEMORIZATION TO CONCEPTUAL MASTERY

The future of accounting training is in teaching key concepts, the core principles that undergird them, and the variables involved in cause-and-effect relationships.

“They have to understand the principle,” Mayes said. “They have to understand the concept of vouching, why you would do it, what an invoice is, what a purchase order is, and what the transaction ledger is, so the concept makes sense to them, and they can supervise a bot that’s doing it for them.”

This need now shapes AICPA strategy.

“We just launched an initiative called ‘Profession Ready,'” Mayes explained. “It’s a holistic initiative that is focused on the early-career skills gaps, with a look toward where the puck is going versus where it is today.”

At the heart of the initiative is a research project focused on the roles and required skills of early-career and aspiring CPAs, with an emphasis on identifying current gaps and anticipating future needs. Mayes expects findings to be finalized and shared in 2027.

AI AFFECTS ALL RUNGS OF THE CAREER LADDER

David A. Wood, Glenn D. Ardis Professor in the School of Accountancy at Brigham Young University in Provo, Utah, believes AI’s impact extends beyond entry-level work. AI already can do certain tasks of a staff member, a manager, and a partner. All levels are being affected simultaneously.

Wood observed this firsthand when building an AI-based Excel tool that automated parts of the review process. Reviewing is typically the job of a senior, a manager, and a partner, he said. But the Excel tool they built is used by entry-level staff, he said. “They can see what the problems are and fix them before sending them to their manager, so they get detailed feedback and improve without requiring training time from their manager.”

Examples like this illustrate why Wood believes AI has the potential to compress learning cycles.

USE AI AS A TRAINING PARTNER

Wood, who is a member of the Profession Ready Initiative Advisory Group for research, also came up with an approach where students teach AI as part of their training.

“Students don’t remember a lot from sitting and listening to a lecture,” he said. “But if they teach somebody else, they remember a ton. So, we thought, what if we had them mentor an AI?”

Instead of reading and testing, students train a deliberately “ignorant” AI until it can pass an exam.

“They read a textbook chapter and then have to train an AI until the AI can pass the quiz,” Wood said.

The approach flips learning from passive to active and mirrors how modern professionals will work with AI in practice. The same approach is being used to train people skills.

“Interviewing is a big thing in accounting, so we built another tool where they interview the AI persona,” Wood explained. “They get practice and feedback on both their [people] skills and interviewing skills.”

THEORETICAL UNDERSTANDING ISN’T ENOUGH

While university education is evolving, firm leaders see a dangerous gap growing between the theoretical understanding graduates have gained about accounting practices and the application necessary in a firm.

“A student will graduate understanding how to do lease accounting from a theoretical perspective,” said Elizabeth Mason, CPA, CEO and founder of High Rock Accounting in the greater Phoenix area, sharing her experience, “but you ask that student to apply that, saying, ‘Go ahead and book these 10 leases,’ and they have no idea how to create the right spreadsheet. They don’t know the process.”

Her solution is to build AI into firm training.

“We could build a RAG (retrieval-augmented generation) on top of any LLM (large language model) utilizing High Rock datasets,” Mason said. “Here are all our reconciliations. Here’s the way we do it. Our team members could then come out of school and ask it any question in the world.”

This approach allows junior staff to deliver advanced outputs quickly, she said, but there’s a hidden danger. While AI has the potential to improve output quality, it also raises the stakes for judgment.

“They may not be able to review it, and that’s the risk,” Mason said. “It’s really about putting in controls at the firm level around a review of work.”

Another major training shift is the need to understand not just what AI says, but how it works.

“A large majority of accountants have no understanding of the architecture or the implications of that architecture,” Mason said.

That’s why she believes future training must include:

• Prompt engineering: Learning how to craft inputs with the right balance of specificity and generality to receive the desired outputs from LLMs. 
• Context design: Understanding the architecture of the systems you’re using and whether they incorporate contextual engineering on the back end so you can determine the appropriate level of review and critical thinking. 
• Data governance: Creating frameworks and processes to keep sensitive data secure and ensure all employees know which data is acceptable to feed into LLMs.
• Model limitations: Understanding that no system is perfect and that AI has inherent limitations when it comes to contextual nuance and a reliance on imperfect datasets. 
• Source validation: Knowing how LLMs source information. LLMs like ChatGPT that use a vector database have a higher risk of hallucinations than systems that are using only a graph database and exact references.

UPSKILLING AT SCALE INSIDE FIRMS

Theresa Richardson, CPA, partner and chief talent officer at Withum, a nationwide firm based in Princeton, N.J., described how her firm is moving to structurally redesign their work.

“We are thoroughly evaluating all the tasks that our team members handle to determine how technology and artificial intelligence can streamline these processes,” Richardson said. “We are also pushing certain responsibilities down from senior to junior team members, creating faster opportunities for hands-on learning and development.”

She emphasized the need to cultivate an “AI analytical mindset” among accounting professionals. “We must train our teams not only to use AI, but to critically assess and identify when AI-driven outputs may be incorrect,” she said.

Beyond technical acumen, Richardson highlighted the importance of developing a consultant’s perspective. Accountants must be able to interpret the results provided by AI and effectively communicate the meaning and implications of this information to clients or to CEOs and boards of directors.

Mayes explained how publicly available LLMs can create hidden compliance risks when teams are not properly trained to use the right tools and validate outputs.

“Say it’s pulling from PCAOB standards, but it doesn’t have access to the Yellow Book,” he said. “You might have a staff approaching an audit in an inappropriate way.”

Mayes argued that firms and finance departments must move toward controlled AI systems trained on verified standards. “If you can train an agent to leverage the AICPA’s entire professional standards suite, then you’ve got something much more valuable,” he said.

Even then, a knowledgeable human must evaluate its output.

THE FUTURE OF EDUCATION MAY BE SIMULATION-BASED

Immersive learning is another training focus. For example, Wood described building a 3D simulation for his students to teach them about auditing inventory.

“We built a full 3D simulation where they go in, open the boxes, and look for internal control failures,” he said. “Rather than simply read about the process in a textbook, it allows them to go around and talk to AI-simulated people, ask for documents, and review them.”

This could be the new version of on-the-job learning, but safer, faster, and scalable.

FROM ‘USING AI’ TO ‘BECOMING WITH AI’

Wood described the future professional mindset as “becoming with AI.” He explained that while some people approach AI as a machine that simply does the job for them, others consider the output and even challenge it, improving both the results and themselves in the process. 

This is the defining difference: AI as a shortcut vs. AI as a cognitive amplifier. Training must explicitly teach the second model.

Despite the dramatic tools available, Wood argues that the real constraint is not AI, but adoption.

“The real bottleneck is the humans in adopting this,” he said. “The tech keeps changing faster than we are able to absorb it.”

This means training will never be “done.” It will be perpetual.

Another reason for resistance to AI adoption likely stems from a fear that technology will replace human work. But Wood believes AI and automation will disrupt the job market in ways we’ve seen before. He pointed out that pretty much every technological revolution has created more jobs than it eliminated. 

“I don’t foresee massive job loss for everybody,” he said. “I do see massive job change for most.”

THE FUTURE OF ACCOUNTANCY TRAINING

For a new era of training to be successful, academia and employers need to collaborate, Wood suggested.

“I have found that firms are right in the same spot we are in terms of learning, understanding, and educating their people,” he said. “This is a unique time in history where everybody starts the race at the same level.”

For Mayes, Wood, Mason, and Richardson, that future of accountancy training must focus on:

1. Conceptual understanding over mechanical execution: Accountants must understand why processes exist, not just how to perform them.
2. Supervision of AI: Professionals must be trained to evaluate, challenge, and govern automated work.
3. Simulation–based learning: 3D environments, AI role-play, and scenario training will replace passive lectures.
4. Cross–level training: Interns and partners must learn together, not separately.
5. Human skills as core curriculum: Communication, consulting, judgment, and trust building will define professional value.
6. Continuous upskilling: There is no stable endpoint, only lifelong learning.

---

About the author

Hannah Pitstick is a content writer for the AICPA and CIMA. To comment on this article or to suggest an idea for another article, contact Jeff Drew at Jeff.Drew@aicpa-cima.com.

---

MEMBER RESOURCES

Articles

“The Accounting Graduate Pipeline: Where Do Things Stand?” JofA, Oct. 27, 2025

“Skilled for Success? Accounting Newcomers Say Yes, Managers Say No,” JofA, Sept. 9, 2025

JofA Artificial Intelligence (AI) Coverage

Podcast episode

“Reflecting on AI’s Rise in Accounting, Looking to What Comes Next,” JofA, Oct. 23, 2025

Website

Profession Ready Initiative landing page

Find the latest information and participate in pulse surveys open to anyone willing to share input about key early-career skills for accountants.

---

LEARNING RESOURCES

Core Concepts of Artificial Intelligence for Accounting Professionals

Unlock the power of artificial intelligence and gain a solid foundation in AI components.

CPE SELF-STUDY

Career Launchpad — Ultimate

The first step toward your dream career is often the most intimidating. Give yourself the best chance of securing a new job by completing the Career Launchpad, which will guide you through potential career opportunities, demonstrate how to market yourself, and prepare you for your job interview.

CPE SELF-STUDY

AICPA ENGAGE

Artificial intelligence and audit will be on the agenda as the biggest event in the accounting profession celebrates its 10th anniversary at the ARIA in Las Vegas.

June 8–11

CONFERENCE

For more information or to make a purchase, go to aicpa-cima.com/cpe-learning or call 888-777-7077.